Journal of Behavioral Medicine

, Volume 31, Issue 3, pp 259–279

Anger inhibition and pain: conceptualizations, evidence and new directions

  • John W. Burns
  • Phillip J. Quartana
  • Stephen Bruehl

DOI: 10.1007/s10865-008-9154-7

Cite this article as:
Burns, J.W., Quartana, P.J. & Bruehl, S. J Behav Med (2008) 31: 259. doi:10.1007/s10865-008-9154-7


Anger and how anger is regulated appear to affect acute and chronic pain intensity. The inhibition of anger (anger-in), in particular, has received much attention, and it is widely believed that suppressing or inhibiting the verbal or physical expression of anger is related to increased pain severity. We examine theoretical accounts for expecting that anger inhibition should affect pain, and review evidence for this claim. We suggest that the evidence for a link between trait anger-in (the self-reported tendency to inhibit anger expression when angry) and acute and chronic pain severity is quite limited owing to a number of factors including a inadequate definition of trait anger-in embodied in the popular anger-in subscale of Spielberger’s Anger Expression Inventory, and a strong overlap between trait anger-in scores and measures of general negative affect (NA). We argue that in order to determine whether something unique to the process of anger inhibition exerts direct effects on subsequent pain intensity, new conceptualizations and approaches are needed that go beyond self-report assessments of trait anger-in. We present one model of anger inhibition and pain that adopts elements of Wegner’s ironic process theory of thought suppression. Findings from this emerging research paradigm indicate that state anger suppression (suppression manipulated in the laboratory) may indeed affect sensitivity to subsequent painful stimuli, and we outline potentially productive avenues of future inquiry that build on this model. We conclude that although studies employing correlational designs and self-reports of trait anger-in have not upheld the claim that anger inhibition affects pain severity, evidence from studies using new models suggests that actually inhibiting anger expression during a provocative event may increase perceived pain at a later time.


Anger inhibition/suppression Acute and chronic pain 

Anger is related to heightened acute and chronic pain sensitivity. Although relationships of pain with anxiety and depression have been widely studied (Banks and Kerns 1996; Keefe et al. 2004), work by Berkowitz (1990, 1993) gives reason to suspect that the connection between anger and pain may be strong and unique, and not completely reducible to the effects of a common substrate of negative affectivity.1 Indeed, findings suggest that anger arousal may be more potent than the arousal of other negative affects in eliciting pain and concomitant physiological reactivity (Burns 2006a). Whereas patients with various chronic pain disorders appear to be characterized by high levels of trait anger and hostility (Hatch et al. 1991; Wade et al. 1990), many findings implicate the manner in which angry feelings are regulated as a particularly robust correlate of acute and chronic pain severity (Kerns et al. 1994; Bruehl et al. 2003a, b; Burns et al. 1996). As in the general emotion regulation literature (e.g., Gross 2002), much attention in pain research has focused on the inhibition or suppression of anger (Keefe et al. 2001).

To better document and understand any links between anger inhibition and physical pain, at least four interrelated questions must be addressed. What is anger inhibition? What reasons do we have for expecting that anger inhibition and physical pain are associated to an important degree? What is the evidence that they are indeed related? To the degree they are associated, are they linked by the mechanisms proposed by theory? We argue in this paper that answers to these questions provided by extant theoretical and empirical accounts are far from adequate, and that new ways are needed to conceptualize relationships between anger inhibition and pain so that understanding may be furthered.

Defining anger inhibition

Even at this late date, it is not entirely clear what “anger inhibition” is. Historically, attempts to define anger inhibition—or “anger-in”—have taken many tacks. Early attempts to systematically examine anger inhibition did so in the service of elucidating so-called psychosomatic phenomena, and were the intellectual offspring of Freud’s notions of repression and conversion (Freud 1957). Examples include Engel’s (1959) depiction of “pain-prone” people who tend to inhibit anger and aggressive impulses, and Alexander’s (1948) description of hypertensive patients as having their anger directed inward toward the ego. Results of these investigations were largely descriptive, and in many respects were more focused on the effects than on the cause. Anger inhibition was but one component of the profiles of “pain-prone” and hypertensive patients, and so was not subjected to rigorous attempts at definition. Typical of this theoretical tradition, as well, the notion of anger inhibition was not codified by assessment instruments—self-report or otherwise—or tied to investigational operationalizations of the construct.

Funkenstein et al. (1954) provided implicit definitions of anger expression and inhibition by classifying subjects into two groups on the basis of their reports of thoughts and emotions they experienced during laboratory manipulations. Anger expressors were so labeled if they indicated that they expressed anger to, or felt anger toward the experimenter, whereas anger inhibitors were identified on the basis of reporting feelings of anger with themselves. The definition of anger inhibition implied by this measure focused on subjects’ feeling angry about their poor performances—anger directed inward toward the self—and was not necessarily concerned with subjects halting verbal and/or physical anger expressions. Harburg et al. (1973, 1979, 2003) defined anger inhibition as acknowledging angry feelings during provocation, but then not showing signs of this arousal or withdrawing in silence. They classified individuals into anger expressor or anger inhibitor groups on the basis of their reports of what they would feel and do in response to vignettes describing anger-provoking situations. Here, we see two issues that still plague us. Both Funkenstein and Harburg, like Engel and Alexander were more focused on the putative effects of anger inhibition than on the constructs themselves. Thus, psychometric evaluations of the instruments were not a priority, leaving us without a good grasp of whether these instruments validly tap anger inhibition at all. Second, we see an important divergence in definitions in that Alexander and Funkenstein saw anger inhibition as a process of directing anger inward toward the self. Engel and Harburg, in contrast, saw it as the actual halting of verbal and physical responses associated with angry emotion, following Newcomb’s notion that anger inhibition can be “observed” by the absence of what should appear during provocation (1950).

The construction of Spielberger and colleagues’ Anger Expression Inventory (AEI; 1985) was undertaken to overcome the difficulty with classifying people on the basis of responses to two or three vignettes, per Harburg, some of which the person may never have experienced, and to disentangle the assessment of how much anger people felt from how they respond to or regulate angry feelings. So-called “anger management style” was conceptualized as a trait or individual difference variable, in that Spielberger and colleagues reasoned that people were possessed of a predilection to respond to a wide variety of anger-arousing situations in one way or another. Consistent with this conceptualization, the AEI was designed so that people respond to a prompt describing a general situation (i.e., “when angry or furious”) by endorsing items tapping the degree to which they typically inhibit (“I pout or sulk,” “I am irritated more than people are aware”) or express anger in different ways (“I say nasty things,” “I slam doors”). Thus, anger-in is defined by this instrument as the tendency for an individual to recognize and acknowledge experiencing anger but then to inhibit verbal and physical expression. Anger-out is defined as the tendency to recognize and acknowledge experiencing anger and to verbally and physically express it.

The conceptualization of anger-in as a trait and the definition of anger inhibition as avoiding or interrupting verbal or physical expression, embodied by the items of this instrument, have come to predominate in empirical psychology. Indeed, the vast majority of empirical studies over the past 20 years addressing some aspect of anger management style, including studies of pain and anger, have used the Spielberger AEI. What we think we know about how and why anger inhibition may be linked to any number of important indexes of physical functioning can be reduced to what we know about associations between scores on the anger-in subscale of the AEI and these indexes. We argue that our present understanding of empirical links between anger inhibition and pain has been shaped by a largely uncritical acceptance of how this construct is defined by the anger-in subscale. Although the definition of anger inhibition offered by the anger-in subscale of the AEI certainly has a pedigree in that it embraces the enduring awareness that some people under at least some circumstances somehow—deliberately or reflexively—avoid or interrupt the verbal and/or physical expression of angry emotions, we must consider some issues of validity that may limit our confidence in relying exclusively on this subscale to exhaustively represent the construct.

One validity consideration is that Spielberger and colleagues expected that anger management style would be a unidimensional construct with pronounced anger inhibition on one end of a continuum and pronounced anger expression on the other. Instead, results of factor analyses revealed that items tapping anger inhibition and anger expression loaded on two separate factors, indicating that the tendency to inhibit anger and to express anger describe two distinct anger management traits. Individuals may vary from “never” to “always” on whether they tend to inhibit anger and whether they tend to express anger. Indeed, far from being negatively related as the original conceptualization held, results across a variety of studies indicate that to the extent the anger-in and anger-out subscales are related, there tends to be a weak but positive association (e.g., Linden et al. 2003; Rene et al. 1999; Riley and Treiber 1989). Put simply, Spielberger appears to have uncovered two fairly distinct constructs of anger management that are only weakly correlated. Because these dimensions are largely orthogonal, one could conceive of “anger management style” as comprised of at least two axes, with one “quadrant” made up of people who score high on both anger-in and anger-out. Thus, high scorers on the anger-in subscale will include people who also score low or high on the anger-out subscale, thus covertly adding a further distinction among anger inhibitors in which those who inhibit under most circumstances are distinguishable from those who inhibit and express in equal measure. We argue that implications for the construct validity of the AEI subscales stemming from the unexpected near-independence of the anger-in and anger-out subscales have not been fully appreciated by researchers. The two implicit anger-in “groups” (i.e., high anger-in/low anger-out and high anger-in/high anger-out), despite sharing high anger-in scores, may differ dramatically on a host of personality, social, and physical health factors as a function of their variability on anger-out.

Second, as we will explore in depth below, evidence suggests that the anger-in subscale may not possess adequate discriminant validity from broadly defined negative affectivity. Findings across numerous studies suggest that the anger-in subscale is moderately related to measures of depressed mood, anxiety and trait anger (e.g., Rene et al. 1999; Spielberger et al. 1985). To the degree that the anger-in subscale is associated with measures of other constructs, its ability to assess a unique anger inhibition construct may be diminished. Worse, as we will see below, given these moderate associations with broadly defined negative affectivity, the anger-in subscale may be related to pain only insofar as it shares variance with these other measures.

A third issue to be considered is that anger-in conceptualizes anger inhibition as a tendency to fully recognize and acknowledge that one is angry. This notion of full recognition of angry feelings may miss an important type or element of anger inhibition that harks back to more psychodynamic conceptualizations; namely, coping with anger through repression (Burns et al. 1999). Recall Weinberger et al.’s (1979) notion of repressive copers who are defined by their low scores on self-report measures of anxiety and high scores on scales tapping defensiveness. These people are believed to disavow negative emotions during noxious events, but experience them nonetheless as revealed by high levels of physiological arousal (Asendorpf and Scherer 1983; Newton and Contrada 1992). Rather than recognizing and acknowledging the experience of anger but then deliberately suppressing it, an anger repressor may deny angry feelings altogether. Because the items on the AEI begin with the prompt, “When angry or furious,” anger repressors will tend not to be identified by items on the anger-in subscale alone due to their tendency to deny ever being angry. Indeed, results of at least one study suggest that individuals who report low anger-out but score high on defensiveness may reveal this classic discrepancy between self-report and physiological reactivity, and then only when they are subject to interpersonal provocation (Burns et al. 1999). Although anger repressors reported levels of angry feelings during harassment comparable to those of low anger expressors, they exhibited heart rate reactivity comparable to high anger expressors. Both people scoring high on anger-in and those scoring low on anger-out but high on defensiveness may therefore appear as anger inhibitors. The former group, however, are thought to be consciously aware of feeling angry but deliberately withhold expression, whereas the latter group may not be consciously aware of their anger and so reflexively curtail behavioral signs of it, even despite experiencing high levels of physiological activation. These, it may be argued, are quite distinct conceptualizations of anger inhibition, and so may have different links to acute and chronic pain.

Fourth, investigators have also argued that anger-in as defined by the AEI simply does not adequately capture the breadth of anger inhibition. For instance, Weber and colleagues (Weber 2004; Weber et al. 2004) suggest that inhibited anger expression may be manifested as the result of three distinct ways of responding to angry feelings: (1) cognitive restructuring, (2) rumination and (3) submission. The last two tactics are defined by Weber as ineffective at reducing anger, and may map well onto the anger-in items of the AEI. Linden and colleagues (Hogan and Linden 2004; Linden et al. 2003) have also identified multiple empirically-distinct dimensions of anger-in. Results of factor analyses showed that anger inhibition may appear as high levels of anger avoidance and anger diffusion, which appear akin to cognitive restructuring and distraction, and as low levels of assertiveness. Their results also pointed to rumination as closely related to anger inhibition. Further, Sukhodolsky et al. (2001) acknowledge the close similarity of anger-in as defined by the AEI and rumination, but suggest that the anger-in subscale tells only of whether a person inhibits anger expression. They argue that the items of the anger-in subscale do not address well enough what happens to anger after it has been suppressed. They submit that rumination may be a complementary process in which anger is revisited, rehashed, and kept unresolved after overt expression is withheld. In short, knowing only whether a person tends to inhibit anger may not be sufficient to understand subsequent effects on psychological and physical health, as behavioral suppression may only be an initial stage of regulation, with subsequent cognitive and emotional “post-event processing,” as described in the social anxiety literature (Dannahy and Stopa 2007; Kocovski et al. 2005) taking place following suppression. Taken together, these findings imply that inhibition of anger may be the observable manifestation of a consistent lack of overt verbal and physical expression following arousal of anger. However, a wide variety of additional processes may underlie and/or follow the inhibition of anger, which may have divergent effects on the maintenance of angry feelings, and so may differentially affect psychological and physical health variables, including acute and chronic pain.

Finally, it is important to consider the broader literature on emotion regulation when considering a definition of anger inhibition (e.g., Gross 2002). Indeed, anger, like all emotions, is comprised of at least three interrelated response channels: physiology, behavior and subjective experience. Inspection of items comprising the anger-in subscale suggests that they tap almost exclusively the behavioral or expressive component of anger inhibition. However, as outlined in Quartana and Burns (2007), suppression efforts can be levied as well at the subjective experiential aspects of anger. Specifically, when angered, an individual may try to put thoughts of the event “out of mind,” or to try to not feel angry during an anger-provoking situation. Attempts to inhibit behavioral or experiential channels of angry emotion may exert distinct patterns of effects on physical health indexes in general, and on pain in particular. It may be the case that any definition of anger inhibition that fails to take into account the inhibition of multiple channels of angry emotion may obscure from view the full range of possible links between this construct and physical pain.

In sum, although our definitions of anger inhibition have moved from clinical impression to measurable constructs grounded in psychometrically sound assessment instruments, the near-exclusive reliance of empirical investigations on the anger-in subscale of the AEI appears fraught with difficulties. That the anger-in subscale does not define an anger-in style that lies at one end of a continuum anchored opposite by an anger-out style, that the anger-in subscale appears to best tap a kind of conscious and deliberate anger inhibition apart from unconscious repression, and that the anger-in subscale targets only one particular channel of anger inhibition and does so via a “snap-shot” approach, all suggest that caution be used in deciding that this subscale adequately taps what we refer to as anger inhibition. Taking these considerations together, it appears that the anger-in subscale may not necessarily be measuring the construct we think it is measuring. Hence, we submit that using the anger-in subscale alone to define anger inhibition may obscure from view it’s most toxic and, conversely, it’s most healthy elements.

Theoretical conceptualizations of links between anger inhibition and pain

We can take away from the above discussion that a common feature of the definitions of anger inhibition implicitly or explicitly supplied by Funkenstein, Spielberger and the others is that anger inhibition, following Newcomb (1950), is manifested as a lack of expected anger behaviors. Even given this very generic description, what reasons are there for expecting anger inhibition to have anything to do with maintaining, exacerbating, or even causing pain? Moreover, what evidence exists to support an empirical association?

Psychodynamic roots

The impetus for examining anger inhibition and pain is largely founded in psychodynamic theory. Freud believed that much mental and behavioral effort is expended to erect defenses against psychologically threatening thoughts and wishes. Repression describes a process whereby anxiety, aroused by a host of beliefs, memories and stimuli, is avoided or reduced by driving the provocative material from conscious awareness (Holmes 1990). Beliefs that expression or even feelings of anger are reprehensible, socially unacceptable or frightening may produce anxiety during events that arouse anger, leading to the inhibition of angry feelings and behavior. Repressed emotions, however, may not remain solely covert mental events. Freud (Breuer and Freud 1955; Freud 1957) advanced the theory of “conversion hysteria” to capture a phenomenon by which people repress alarming and unacceptable emotions or impulses to act in an effort to reduce anxiety, but also transform arousal attending the threatening appraisals or stimuli into physical symptoms. The notion that a variety of physical symptoms arising from repressed or inhibited emotions could appear in the absence of organic pathology inspired a number of theories and studies about the nature of chronic pain.

For instance, Szasz (1957) held that chronic pain may help some people deal with catastrophic emotional turmoil, but only insofar as it allows them to cope indirectly with distress by seeking treatment for physical symptoms. Engel (1959) proposed that certain “pain-prone” people are more likely than others to employ pain as an “emotional regulator” that assuages guilt, replaces loss, or appears as a manifestation of inhibited anger or aggressive urges that the individual will not acknowledge or act on. These theories spring from observations that chronic pain is often expressed far in excess of demonstrable pathology, and tends to arise in people who appear also to have difficulty managing and expressing strong negative emotions. Thus, chronic pain may allow some patients to endure, explain, complain about, and seek help for symptoms that are not psychologically threatening, but which are inextricably intertwined with the suppression and inhibition of emotional agitation they can neither accept nor express directly. Compelling though they may sound, such theories do not easily lend themselves to empirical scrutiny replete with operational definitions. Indeed, these accounts of emotional inhibition and pain never progressed much past their origins in clinical observation and anecdote.

A number of contemporary theories borrow at least implicitly from these psychodynamic formulations. These models, however, provide propositions that are amenable to empirical tests regarding why we may expect inhibited anger and pain to be linked. We outline some of these models below.

Negative affect and magnification of physical symptoms

People characterized by a tendency to inhibit the expression of anger may report elevated pain intensity or sensitivity to noxious stimuli because they also amplify their perception of somatic symptoms as a function of a general tendency for them to experience negative affectivity and emotional distress. Watson and Pennebaker (1989) argued that trait NA—the tendency to experience high levels of distress, dissatisfaction, and negative emotions at all times—and the tendency to report many physical symptoms are inextricably intertwined. A proneness to subjective distress extends across negative mood, pessimistic outlook and sensitivity to physical symptoms. As argued by others (Brosschot 2002; Janssen 2002; Raphael et al. 2000), one way to understand the link between NA and increased physical symptom reporting is through the notion of somatosensory amplification. All distressing and threatening aspects of life, including physical symptoms, are selectively attended to, and magnified in importance and potential danger. Evidence suggests that high NA individuals do not necessarily have greater sensitivity to physical stimuli, as revealed during signal detection paradigms (Aronson et al. 2006), but report more symptoms due to a response bias to physical sensations.

To the degree that anger inhibition, as tapped by self-report scales like the anger-in subscale of the AEI, is related to constituents of broad NA—that is, to self-reports of depression, anxiety, anger—it too may be associated with high levels of physical symptom reporting. Thus, we could expect anger-in to be related to pain not because of some special quality of inhibiting the expression of anger, but because anger-in shares variance with an underlying trait that influences pain perceptions through distress-driven symptom amplification.

Inhibition of thought, emotion and behavior

Theories of inhibition focus closely on the unique effects of inhibition, maintaining in general that suppressing awareness of, thoughts about, and/or the physical expressions inherent in strong (negative) emotions may have adverse consequences for psychological and physical health. In general, these theories hold that suppression or inhibition of strong emotion is an active, effortful undertaking in which inhibitory processes are recruited and then pitted against ongoing cognitive or behavioral activity. Such inhibition exacts psychological and physiological costs. This central tenet has been pursued, and to some extent, validated through laboratory and field investigations.

Gross and colleagues (Gross 1999; Gross and Levenson 1993, 1997) have examined effects of antecedent- and response-focused emotion regulation on subjective and physiological responses to emotion induction. The former strategy involves cognitive reappraisal of a stressful event prior to the experience of full-blown emotion, whereas the latter involves suppression of expression during the experience of an ongoing emotional response. Their findings indicated that expressive suppression increases sympathetic nervous system activity in response to emotion-eliciting film clips to a greater extent than that revealed by people allowed to react to emotion-inducing stimuli without interference. Results also showed that participants in the expressive suppression conditions revealed less outward displays of emotion than control participants, although suppression did not appear to reduce emotional experience. These findings speak to the “effort” purportedly required to effectively inhibit or suppress outward expressions, but also underscore the distinction between behavioral and experiential aspects of emotion regulation. Moreover, results of additional studies suggests that people who habitually suppress expression of emotion fare worse on a number of psychological and physical health indicators than people who typically regulate emotion through cognitive reappraisal (Gross 2002; Gross et al. 2000). To the extent that anger inhibition may be part of a broad tendency to suppress emotion expression in general (Burns et al. 2007), we may expect anger-in to also exert ill effects on physical health factors, such as acute and chronic pain severity particularly through effects on physiological variables (e.g., symptom-specific muscle tension, Burns et al. 2006; see below).

Rationale for expecting anger inhibition to affect acute and chronic pain may also be provided by theory and findings regarding the disclosure of emotional events. Prolonged avoidance and repression of stressful or traumatic events appears to prevent the adaptive reappraisal, integration and settling of such experiences (Greenberg, and Safran 1987; Pennebaker 1989). Investigators propose that the suppression of emotions evoked during stressful and traumatic events requires effort and so exacts psychological and physiological “costs” (e.g., Pennebaker 1989). The disclosure of such events would reduce or eliminate the felt need to suppress them, and so ameliorate adverse effects traced to the exertions required to inhibit. Put simply, if inhibition requires costly effort, then “undoing” it should provide psychological and physiological relief. Pennebaker and colleagues found that subjects who wrote about previously unacknowledged traumatic events exhibited a decrease in subsequent health center visits (Pennebaker and Beall 1986), and enhanced immune function (Pennebaker et al. 1988) compared to no-writing control participants. Results of Greenberg and Stone (1992) suggested that the beneficial effects of disclosure were a function of a “dose-response” of the amount of emotion disinhibited during disclosure. Results indicated that subjects disclosing severe traumas showed greater decreases in later physician visits than subjects disclosing mild traumatic events.

Most pertinent to the present review, results of recent research using the written disclosure paradigm suggest that the disinhibition of negative emotion has salutary effects on pain and chronic pain-related difficulties. Sullivan and Neish (1999) randomly assigned dental patients to either disclose their thoughts and feelings about dental treatment or to write about their activities of the previous day. They found that patients who disclosed had lower pain and distress following their dental procedure than control subjects, although these effects were most conspicuous among patients in the disclosure condition who tended to catastrophize about pain. Kelley et al. (1997) randomly assigned rheumatoid arthritis patients to either speak into a tape recorder about stressful or traumatic events or to give descriptions of neutral pictures. Patients in the disclosure condition reported less affective disturbance and physical dysfunction at three-month follow-up than control patients. Similarly, in two controlled outcome studies (Broderick et al. 2005; Gillis et al. 2006), investigators found that fibromyalgia patients assigned to written disclosure conditions had better long-term outcomes (3–4 months post-writing) than control patients.

Although some null effects have been reported for clinical pain populations (for review, see Frisina et al. 2004), findings generally extend links between disinhibition and general health to the pain arena. Disinhibiting suppressed thoughts and emotions about traumatic or stressful events appears to alleviate some of the physical suffering associated with acutely painful procedures and longstanding conditions, suggesting that inhibition of emotion expression does indeed play some role in the exacerbation of pain severity, physical symptoms and the disability that attends them. As with the expressive suppression paradigm of Gross and colleagues, we can expect the inhibition of anger to be related to pain insofar as anger inhibition is an example of suppressed thought and emotions surrounding a stressful event. Of note, it may be the case that disclosure of intense but inhibited anger, particularly when anger was a defining reaction to an event, could produce the same kind of benefits as disclosure in general. To our knowledge, such results have not been reported.

We have argued that one theory in particular may be useful in trying to understand links between anger inhibition and pain (Burns et al. in press; Quartana and Burns 2007; Quartana et al. 2007). Like Gross’ paradigm, the ironic process theory of mental control developed by Wegner and colleagues (Wegner 1994; Wenzlaff and Wegner 2000) focuses on deliberate attempts to suppress, but unlike Gross, the latter investigators also explicitly attend to effects that follow suppression. Also unlike Gross, the work of Wegner begins with the mechanisms involved in attempts to inhibit or suppress thoughts rather than the, perhaps, subsidiary effects of inhibiting behaviors. We believe that Wegner’s ironic process theory may shed light on how anger inhibition at one point can affect pain severity at a later point.

Wegner and colleagues (Wegner and Erber 1992; Wegner et al. 1990, 1993; Wegner and Gold 1995) showed that attempts to suppress unwanted thoughts in order to gain mental control have the ironic effect of making these thoughts more insistent and salient. They proposed that efforts to suppress thoughts involve the interaction of two processes (Wegner 1994). First, an intentional, effortful “operating process” works to avoid unwanted thoughts through controlled and conscious use of distracters. A second process, however, occurs that is relatively resource-independent, less conscious, and which begins when suppression is initiated. This “monitoring process” searches for mental contents that signal a failure to avoid the unwanted material, thereby paradoxically enhancing accessibility of the undesired material. Wegner et al. describe two kinds of thought suppression effects. Immediate enhancement effects are those that arise during suppression, and are reflected in a higher frequency of reported intrusions or increases in physiological activity during suppression. Rebound or carryover effects occur when suppression is terminated, and are also reflected in increased frequency of thoughts about the suppressed material, or sustained physiological arousal. Results of a recent meta-analysis suggested that rebound is more robust and reliable than immediate enhancement effects (Abramowitz et al. 2001). Much research in a variety of populations supports the ironic process model (Newman et al. 1997; Petrie et al. 1998; Tolin et al. 2002).

The theories of Pennebaker, Gross and Wegner and results of empirical work associated with them suggest that factors unique to the process of suppression/inhibition lead to adverse psychological and physical health problems.


For the somatic symptom amplification model, anger-in as a trait exerts effects to the extent that other processes exert effects. In this conceptualization, anger-in is a component of the tendency to experience NA, and so it does not allocate a central role to processes unique to the active inhibition of anger expression. Thus, anger-in may predict acute and chronic pain intensity because of factors and mechanisms that have little to do with inhibition per se. The models of Pennebaker, Gross and Wegner do focus explicitly on inhibition processes, but do not target anger inhibition in particular. These theories are concerned with how traumatic, stressful and even day to day negative emotional material is handled, which may involve suppression of manifold negative thoughts and emotions. That is, these are generic models of suppression and inhibition. Indeed, recent studies have examined how suppression of pain may have adverse effects on responsivity to subsequent stimuli that invoke both unpleasant physical sensations and psychological stress (Cioffi and Holloway 1993; Burns 2006b; Elfant et al. in press). The gist of these inhibition theories is that in attempting to suppress unwanted thoughts and feelings or to inhibit certain behaviors, the opposite effect may be achieved; a paradoxical state of affairs that may pose grave consequences for cognitive, emotional and physical functioning as further events unfold.

The somatic symptom amplification and inhibition theories supply different reasons for expecting that the inhibition of anger may be related to acute and chronic pain intensity. But to what extent does the evidence support connections between anger-in and acute and chronic pain intensity? And to what extent are these various conceptualizations for expecting anger-in to be related to pain supported by evidence?

Empirical evidence for links between trait anger-in and pain

As mentioned above, the vast majority of empirical reports on whether anger inhibition is related to acute and chronic pain intensity have relied on the anger-in subscale of Spielberger’s AEI. Most of what we know about effects of anger inhibition on pain is based on this self-report trait measure. Moreover, most of these studies have used cross-sectional, correlational designs.

Studies comparing pain patients to healthy controls on trait anger-in

One approach to the study of relations between trait anger-in and pain has compared patient groups to healthy, pain-free participants. These studies address the proposition that if anger inhibition somehow causes chronic pain, then patients with such disorders should report inhibiting anger more than other groups. At first glance, results of several studies comparing patients with pain conditions to healthy controls seem to indicate that the former report a significantly greater tendency to inhibit anger than the latter. Four studies of headache patients (Arena et al. 1997; Hatch et al. 1991; Materazzo et al. 2000; Nicholson et al. 2003), two studies of fibromyalgia patients (Amir et al. 2000; Sayar et al. 2004) and one study of chronic pelvic pain patients (Thomas et al. 2006) reported statistically significant differences, whereas one study of headache patients (Ham et al. 1994) and one study of low back pain patients reported nonsignificant differences (Bruehl et al. 2002a). Effect sizes, comparing pain groups to healthy controls, reveal a wide range, including negative relationships (d = −.34 to .85), with an average effect across studies suggesting at least a modest difference between groups (mean d = .40). In addition, in an early study not employing the anger-in subscale of the AEI, Pilowsky and Spence (1976) reported that more chronic pain patients reported a tendency to “bottle up anger” than nonpatients. However, in three of these studies, differences between patient groups and controls were not uniform. In Materazzo et al. (2000), cluster-type headache patients differed significantly from healthy controls on trait anger-in, whereas migraine patients did not. In Sayar et al. (2004), fibromyalgia patients reported greater trait anger-in than healthy controls, whereas the rheumatoid arthritis patients did not. Finally, in Grothgar and Scholz (1987), although migraine patients revealed fewer anger expressive behaviors than healthy controls, the low back pain patients did not. Thus, the “box score” considering all reported comparisons is 9 to 5 in favor of a significant effect.

Some considerations sway us from drawing firm conclusions about whether greater anger inhibition distinguishes pain patients from healthy people. First, patients reported significantly higher depression than healthy controls in all of the six studies (Bruehl et al. 2002a; Ham et al. 1994; Hatch et al. 1991; Materazzo et al. 2000; Nicholson et al. 2003; Sayar et al. 2004) that reported these data (range: d = .38–1.28; mean d = .83). Unlike the findings for trait anger-in, all of these comparisons were statistically significant, and appear to represent a larger difference between patients and controls on depression than on anger-in. These findings raise the issue based on the somatic symptom magnification model discussed previously that chronic pain patients may have elevated NA in general, and so links between anger-in and pain may be accounted for by this third variable. This concern is tempered slightly by the finding, in the single study to report such analyses (Nicholson et al. 2003), that the difference between pain and healthy groups on anger-in remained significant after statistically controlling depression. Still, the effect size for the difference between pain group and healthy controls on anger-in in this study was reduced from d = .79 to d = .38 with depression controlled; that is, a 50% decrease.

Second, the majority of studies finding significant differences between patients and controls were concerned with headache patients. In the four studies that included low back pain patients, only Pilowsky and Spence (1976) reported significant differences on anger-in between patients and controls, whereas the others did not (Bruehl et al. 2002a; Grothgar and Scholz 1987; Ham et al. 1994). This disparity across pain disorders prompted us to conduct additional analyses of published data sets (Burns 2006a, b) in which low back pain patients and healthy controls were compared. Results comparing 94 patients to 76 controls and 105 patients to 88 controls revealed nonsignificant differences on anger-in subscale scores [F’s < 1; effect sizes d = −.13 and −.04]. To the extent that high levels of anger inhibition do characterize patients suffering from persistent pain, the effect may be most reliable among groups described by certain pain disorders with particular etiologies. Specifically, for reasons yet to be determined, headache patients may score higher on trait anger-in than healthy controls, whereas low back pain patients may not.

Studies of the association between trait anger-in and indexes of acute and chronic pain

A second approach to the study of trait anger-in and pain is to examine associations among trait anger-in and indexes of acute and chronic pain among both healthy controls and pain patients. These studies tested the broad hypothesis that if anger inhibition influences pain intensity, then there should emerge a consistent and unambiguous statistical association between these factors. Results of 11 studies that examined correlations between the anger-in subscale of the AEI and indexes of chronic pain severity suggest that anger inhibition is not reliably related to level of chronic pain. In five studies of headache patients (Duckro et al. 1995; Ham et al. 1994; Materazzo et al. 2000; Tschannen et al. 1992; Venable et al. 2001) and five studies of samples comprised largely of low back pain patients (Bruehl et al. 2002a; Burns et al. 1996; Conant 1998; Gaskin et al. 1992; Lombardo et al. 2005), the correlations between anger-in and pain severity were nonsignificant, whereas in Sayar et al. (2004, reporting on effects for both fibromylagia and rheumatoid arthritis patients) and in Kerns et al. (1994, reporting on low back pain patients) correlations were significant. Effect sizes ranged from r = −.02 to r = .41 with a mean of r = .18, indicating a very modest effect at best. Further, analyses conducted on our published datasets (Burns 2006a, b) for purposes of this review revealed correlations of r = .01 and −.02 between anger-in subscale scores and chronic pain severity.

However, in three studies (Bruehl et al. 2002a, 2003a, b), investigators reported results for the McGill Pain Questionnaire (Melzack 1987) sensory and affect pain descriptor subscales, rather than for pain severity ratings alone. Here, anger-in scores were significantly correlated with affect descriptor subscale scores in all three studies (range: r = .24–.47; mean r = .31), whereas anger-in scores were related significantly to the sensory descriptor subscale in only one study (range: r = −.01 to .42; mean r = .19). Taken together, findings imply that the tendency to inhibit anger may not affect the absolute intensity of chronic pain to any great extent, or may not even magnify the appraisal of its sensory qualities. Instead, anger inhibition may primarily amplify the negative affective elements of pain.

Again, among all the studies examining links between anger-in and pain intensity, indexes of NA, when reported (Bruehl et al. 2002a; Burns et al. 1996; Duckro et al. 1995; Ham et al. 1994; Kerns et al. 1994; Materazzo et al. 2000; Sayar et al., 2004; Tschannen et al. 1992; Venable et al. 2001), were related significantly to pain severity and/or the sensory and affect descriptor subscales, with effect sizes ranging from r = .11 to r = .51 (mean r = .33). This moderate effect is again somewhat larger than the relationship of trait anger-in with pain. Further, in the seven studies to report these data (Burns et al. 1996; Conant 1998; Duckro et al. 1995; Materazzo et al. 2000; Sayar et al., 2004; Tschannen et al. 1992; Venable et al. 2001), all reported significant correlations between anger-in subscale scores and indexes of depressed mood, with effect sizes ranging from r = .37 to r = .72 (mean r = .54). It is no surprise, therefore, that when analyses were conducted to statistically control depression scores in two of the three studies examining the affect descriptor subscale (Bruehl et al. 2003a, b), the unique effects of anger-in on these descriptors were revealed to be virtually nil (sr’s = .0). As with findings regarding comparisons of pain patients to healthy controls, the notion that anger-in is related to pain mostly because it is also related to broad NA appears to receive support.

Investigators have also examined the relationship between trait anger-in scores and indexes of laboratory-induced acute pain (Bruehl et al. 2002a; Burns et al. 2004, 2007; Gelkopf 1997). Three of four studies reported nonsignificant correlations between anger-in scores and acute pain severity. Only Gelkopf (1997) reported a significant relationship of anger-in to acute pain report (r = .60) Effects ranged from r = .06 to r = .60, with a mean r = .25. Excluding the rather exceptional effect of Gelkopf, which was based on a sample of 21 healthy normals, the mean effect for the remaining three studies was r = .13. In the Burns (2006b) data set, unpublished data revealed the correlation between anger-in scores and acute pain severity to be r = .14 (p > .10). Again, even including the result from Gelkopf, the effect size is no more than modest. However, in two studies (Bruehl et al. 2002a; Burns et al. 2003), investigators examined correlations between anger-in scores and the affect and sensory descriptor subscales of the McGill Pain Questionnaire as used to rate laboratory-induced acute pain. Results showed that in both studies affect descriptors were correlated significantly with anger-in scores (r’s = .43 and .39), whereas sensory descriptors were related significantly to anger-in scores in Bruehl et al. (2002a; r = .33), but not in Burns et al. (2003; r = .18). In both these studies, depression scores were also related significantly to affect descriptor ratings (r’s = .40 and .29). However, Burns et al. (2003) examined the link between anger-in and affect descriptor scores controlling for depression scores, and found the unique effect of trait anger-in to be significant (sr = .29; p < .05). These few results are consistent with findings for chronic pain, and suggest that trait anger-in may be related most reliably to the magnification of the negative affective aspects of acute pain. However, in only one study was the competing explanation of broad NA tested, and at this point we have little reason to suspect that results of additional analyses would not, by and large, parallel what was found when depression was controlled for affective descriptors of chronic pain.


Taking together results from studies that compared pain patients to healthy controls and studies that examined associations between trait anger-in scores and indexes of acute and chronic pain, the evidence overall suggests a weak association between the tendency to inhibit anger and pain intensity. This conclusion runs contrary to the popular psychosomatic hypothesis that inhibition of strong negative emotions leads to poor psychological and physical health.

Where a significant association between trait anger-in and pain was noted, we encounter the additional difficulty of strong covariation with NA. The few studies that actually evaluated the role of this “third variable” do not favor the idea that anger-in, as assessed by the Spielberger anger-in subscale, accounts for much unique variance in pain severity. Trait anger-in may be related to pain intensity mostly to the extent it is associated with the symptom magnification that accompanies high NA. That is, available findings do not implicate any distinct qualities of the process of inhibiting anger in influencing pain, but appear instead to highlight the link between trait anger-in and the symptom magnification that attends distress proneness in general.

It should also be considered that studies to date of links between trait anger-in and pain represent preliminary stages of investigation. Comparing chronic pain groups and healthy controls and examining correlations among measures in cross-sectional, correlational designs can only establish covariation. We cannot rule out, for instance, whether the experience of long-term persistent pain underlies the development of an anger inhibition style. Making firmer statements regarding whether anger inhibition actually causes pain will require longitudinal and experimental designs. Too, even within the limits of correlational designs, very few extant studies attempted to use statistical techniques to gain some control over extraneous variance. In particular, few investigators appeared to pay attention to the problem of the substantial overlap between trait anger-in and NA, and the very real possibility that trait anger-in and pain are linked via somatic symptom magnification.

Empirical evidence for links between state anger inhibition and pain

We (Burns et al. in press; Quartana and Burns 2007; Quartana et al. 2007) have argued that what is missing most from the current literature focused on trait anger-in and pain is the evaluation of the effects of actual inhibition of anger expression on pain. The key proposition here is whether the actual inhibition of anger during an anger-provoking event has any effects at all on sensitivity to or perceptions of subsequent episodes of pain. We looked to inhibition theories to help us develop a theoretical and methodological framework.

Specifically, we borrowed heavily from Wegner’s (1994) ironic process theory of mental control (see above). In short, the ironic process model of anger suppression and pain holds that efforts to suppress experiential and/or expressive components of anger may render this material quite salient and accessible to consciousness as the automatic monitoring system operates to find more and more unwanted anger-related thoughts, feelings and behavioral urges (i.e., testing whether suppression processes are working effectively). Although suppression may subdue anger-related thought, feeling and behavior initially, it may paradoxically lead to increased accessibility of anger-related material over time. Increased accessibility of anger may, in turn, influence or “contaminate” appraisals of subsequent pain with heightened feelings of frustration, irritation and annoyance. We suspect as well that angry responses to the pain itself will be magnified (per the work of Berkowitz). In sum, anger suppression during provocation may cause greater pain severity later as a function of ironically increased accessibility of the to-be-suppressed anger; intrusions that in turn color perceptions of, and emotional responses to later pain—itself a stimulus hypothesized to have subcortical ties to anger and aggression (e.g., Berkowitz 1990; Whalen et al. 2001). In adapting Wegner’s theory of the ironic effects of thought suppression (Wegner 1994), we hoped to evaluate the effects of what are alleged to be the unique elements of the anger inhibition construct; namely, the process and effects of actual (“state”) suppression of anger expression. To do so, we had to move beyond the trait anger-in measures and correlation designs to manipulate anger inhibition and pain-induction in controlled laboratory conditions.

Although suppression of thoughts and behaviors that are part of other negative emotional reactions (anxiety and sadness) may work similarly to affect responses to later stimuli, we speculate that inhibition of anger may produce the most pronounced delayed effects. If anger involves a unique relationship with pain, distinct from those of other negative emotions, per Berkowitz (1990), then we would expect highly accessible anger-related thought, feeling and behavioral inclinations to exert greater contamination effects on pain perception than, for example, sadness-related content. Indeed, research on judgment and perception suggests that highly accessible, or primed, emotions that share emotional overtones with anticipated events will enhance perceptions of that event in terms of the primed emotions (e.g., DeSteno et al. 2004; Rucker and Petty 2004; Schwarz and Clore 1996).

Indirect evidence for the ironic process model of anger suppression and pain has been reported. Cioffi and Holloway (1993) and Elfant et al. (in press) found that healthy normals instructed to suppress pain-related thoughts during a cold pressor showed evidence of rebound effects. Specifically, they reported higher levels of acute pain following termination of the cold pressor and showed greater physiological arousal than subjects assigned to regulate emotion differently (distraction, sensory-focus). Importantly, those told to suppress rated a later innocuous physical stimulus (i.e., a massage device applied to the neck) as more unpleasant and showed greater physiological reactivity to it than the other groups; results that speak to the ability of suppression to exert delayed effects on later events. Further, in Burns (2006b), chronic low back pain patients told to suppress pain-related thoughts during a cold pressor showed evidence of delayed contamination effects during a subsequent mental stressor. Patients instructed to suppress evinced elevated lower paraspinal muscle tension and SBP levels from cold pressor to mental stress to a recovery period, whereas patients regulating emotion differently did not show such reactivity. Taken together, findings indicate that attempts to suppress emotion-related thoughts inspired by unpleasant events may result ironically in their persistence, maintain emotional arousal, and contaminate interpretations of and responses to later events.

Direct evidence for the ironic process model of anger suppression and pain comes from three studies using healthy controls, and one study using chronic low back pain patients. In two of the studies (Quartana et al. 2007; Study 1 and Study 2), healthy participants were criticized while giving a speech, during which they were asked to suppress emotional experience or to react as usual. They then underwent a cold pressor pain task. Results of Study 1 and Study 2 were largely parallel. Participants in the suppression conditions reported greater pain intensity during the cold-pressor and provided greater ratings of perceptions of anger-specific dimensions of the pain (e.g., irritating) than participants in the control conditions. Suppression and control participants, however, did not differ on sensory (e.g., throbbing) or general distress (e.g., fearful) elements of pain. Finally, differences between suppression and control conditions on pain intensity during the cold-pressor were largely accounted for by group differences on anger-specific pain descriptors. That is, magnified perceptions of the irritating and annoying qualities of pain partly explained suppression participants’ greater report of pain intensity.

In these two studies, only anger was induced, and so we could not disentangle effects of suppressing negative affect in general from suppression of anger in particular. Quartana and Burns (2007) examined whether suppression of anger-related experience or expression during mental arithmetic with harassment had effects on pain during the cold pressor that were greater than, and distinct from those of suppression of anxiety-related experience or behavior and a control condition. Here, anxiety was induced by the experimenter silently watching the participant, and checking “answers” on a clipboard. In brief, results showed that subjects who suppressed experiential aspects of emotion and who suppressed expressive aspects of emotion during anger-induction reported significantly greater pain intensity and more strongly endorsed the anger-specific pain descriptors than subjects who suppressed during anxiety-induction, and those who underwent anger-induction but did not suppress. Further results suggested that the degree to which anger suppression groups differed from other groups on sensory and emotionally distressing descriptors of pain was due to their elevated anger-specific references to the cold pressor pain. The degree to which anger suppression affected appraisals of sensory and generally distressing elements of pain was reducible to the unique effect of high levels of anger-specific perceptions. The degree of physical hurt experienced by participants who suppressed during anger induction appeared to be due in part to suppression-induced contamination by anger-related content.

This study replicated and extended the studies in Quartana et al. (2007). Not only did subjects who suppressed experiential or expressive aspects of emotion during anger-induction (i.e., anger suppression) report greater pain severity than those who were angered but did not suppress (as in the two previous studies), those who suppressed during anger-induction also reported greater pain than those who suppressed during anxiety-induction. Hence, results suggest that experiential and expressive suppression in general did not heighten anger and pain responses to subsequent painful stimulation, but that the suppression of anger in particular did.

A fourth study used chronic low back pain patients (Burns et al. in press), the goals of which were to determine whether anger suppression leads to greater subsequent pain report and observable pain behaviors during a task that can produce clinically-relevant low back pain among patients by mimicking everyday activities. Patients were randomly assigned to suppression and nonsuppression conditions, and participated in a computer maze task while being harassed by a confederate. After the maze task, they were taken to another room to undergo a structured pain behavior task based on the behavioral observation procedure described by Keefe and Block (1982). In short, patients in the suppression condition reported significantly greater pain intensity and exhibited more pain behaviors during the pain behavior task than nonsuppression patients. The degree to which anger was aroused during the maze task accounted for (i.e., mediated) group differences in pain behaviors, whereas anxiety and sadness aroused by the task did not. Results replicate previous findings regarding effects of anger suppression, but extend them to reveal suppression-induced increases in clinically relevant pain among a clinical sample.

Critique and summary

Results of these studies strongly suggest that the actual suppression/inhibition of anger—so-called “state” suppression—is related to heightened pain intensity during a subsequent painful stimulus. Because the study designs used experimental manipulation of emotion suppression followed by induction of pain, causal inferences drawn from these findings are more valid that those drawn from the correlational designs used by most studies of trait anger-in. Put simply, results strongly suggest that the actual suppression of anger causes increased pain intensity. Moreover, evidence suggests that a mechanism derived from Wegner’s ironic process theory may connect anger suppression and pain. Namely, to varying degrees, results of the four studies hint that the paradoxical enhancement of accessibility of the to-be-suppressed material—anger—may have exerted delayed effects on pain.

The effect sizes comparing state anger suppression conditions to control conditions on pain intensity showed a range from d = .50 to d = .91, and a mean d = .72, which stand in contrast to the substantially smaller effect sizes found for trait anger-in and pain. Because the associations between trait anger-in and pain reported in the literature appear to be well-explained by somatic amplification and shared variance with NA, it may be that an inhibition model may not be fully applicable to the construct tapped by trait anger-in measures. The evidence that people reporting high trait anger-in actually suppress thoughts and behaviors when angry in ways resembling the processes described by Wegner and Gross is currently very slim, and we cannot at this point rule out the possibility that state anger suppression and trait anger-in are independent phenomena.

Overall, at present, we seem to have two divergent sets of findings. Although the items of the anger-in subscale seem to describe processes by which people suppress/inhibit anger expression, it is not clear whether suppression a la Wegner and Gross are the tactics actually used in everyday life by high anger inhibitors to regulate anger. Suppression may be one of many tactics used to “keep anger in” that, as argued by Weber (2004), are not very effective in reducing anger or its detrimental effects on subsequent events. Importantly, these apparently discrepant findings between trait anger-in and state anger suppression should not be taken as reason to abandon the former in favor of the latter. Instead, we should take a step back to reconsider our conceptualization of trait anger inhibition in light of the issues raised above and findings for actual anger suppression. Again, it is not clear that trait inhibitors as defined by the anger-in subscale actually inhibit or express in ways described by other investigators. Indeed, of the studies concerned directly or indirectly with the construct validity of the anger management subscales, only a handful have included any kind of behavioral or observational measures of actual expression (Burns et al. 1999; Engebretson et al. 1989; Faber and Burns 1996; Lai and Linden 1992). Not surprisingly given the limited ability of trait measures to predict behavior without aggregation across time and place (Epstein 1980), results have been inconsistent. Coming full circle, our definition of anger inhibition still needs considerable attention.

New directions

Mismatch situations

Brosschot and Thayer (1998) have argued that anger suppression may play a larger role in producing or exacerbating physical disorder than once thought because of social norms that discourage full expression. State anger suppression may have the most far-reaching effects among people who typically do not use such an anger regulation tactic, but, through force of circumstances end up doing so. That is, dispositional anger expressors (i.e., high trait anger-out) may find themselves in situations that are mismatches for their anger regulation style. Evidence supports so-called mismatch models with regard to cardiovascular recovery following provocation. Engebretson et al. (1989) found that high anger-out subjects allowed to express anger (in writing) following provocation showed cardiovascular recovery, whereas high anger-out subjects not allowed to express evinced delayed recovery. Faber and Burns (1996) found that greater verbally expressed anger during provocation was associated with faster cardiovascular recovery among high trait anger-out subjects than among low anger-outs.

What about mismatch effects for pain? Results of Burns et al. (2007) examined whether trait anger-out moderated effects of emotion suppression on later pain intensity. In brief, we found that trait anger-out was not related significantly to pain intensity among subjects who either did or did not suppress experiential and expressive aspects of emotion during anxiety-induction. However, trait anger-out was related significantly to a delayed pain recovery (i.e., high pain intensity persisting following termination of a cold pressor) among subjects who suppressed during anger-induction, whereas trait anger-out was not related significantly to pain intensity among subjects who underwent anger-induction but did not suppress. Results of this study suggest that pain sensitivity may be increased among people with high trait anger-out if anger expression is not only blocked in general, but when these people engage in the deliberate suppression of anger in particular.

Thus, state anger suppression may disproportionately affect certain individuals. For people who prefer to express anger, suppressing anger under certain circumstances may lead to greater susceptibility to later painful stimuli and other noxious stimuli. Anger expressors who find themselves chronically unable to express due to, for example, their work environment, may run the risk of developing psychological and physical health problems to a greater than anger expressors having an environment that is a better “fit” to their regulation style. A fuller understanding of how, under what conditions and among whom in particular state anger suppression detrimentally affects physiological reactivity, mood and pain sensitivity appears to be a good target for future research endeavors.

Physiological mechanisms

Physiological mechanisms may play a role in mediating effects of anger inhibition on pain intensity. Several candidates exist. Results of studies by Gross and colleagues (Gross and John 2003; Gross and Levenson 1993, 1997) indicate that sympathetic nervous system reactivity is increased among subjects who use expressive suppression during exposure to stimuli invoking negative emotions. Similarly, in Quartana and Burns (under review), we found that suppression during negative emotion induction was related to greater cardiovascular reactivity than nonsuppression during emotion induction, and that suppression also exerted delayed effects on reactivity during a subsequent pain-induction task. The relationship between sympathetic nervous system reactivity and pain sensitivity is not clear. Findings suggest that, insofar as sympathetic nervous system reactivity is associated with stress exposure, then such reactivity is linked to stress-induced analgesia (al’Absi et al. 1996; Edwards et al. 2006; France and Stewart 1995; Vassend and Knardahl 2004). However, results of other research indicates that increased pain sensitivity is linked to greater sympathetic nervous system activity (e.g., Caceres and Burns 1997; Green et al. 2006; Nilsen et al. 2000), and still other research suggests that both stress-induced sympathetic reactivity and increased pain sensitivity signals a common deficit in functioning of the endogenous opioid system (e.g., Bruehl et al. 2002b; McCubbin and Bruehl 1994). At this point, we at least have enough circumstantial evidence to pursue investigation of sympathetic nervous system reactivity as a mediator linking anger inhibition to pain sensitivity, but keeping in mind that the operation of different neurotransmitters may be responsible for producing such observable reactivity, which in turn may be affected by individual differences and by the different situations in which reactivity is invoked. For instance, we found that SBP reactivity was linked to subsequent pain sensitivity only when subjects first underwent anger-induction (Caceres and Burns 1997), and that this effect was further moderated by individual differences in trait anger-out (Burns et al. 2004).

With regard especially to chronic pain, symptom-specific, stress-induced muscle tension reactivity may represent another physiological mechanism. Investigators speculate that physical and psychological stress may lead to frequent and intense, or low level but sustained muscular contractions (Sjegaard et al. 2000). This muscle activity can increase pain through ischemia and hypoxia (Fields 1987), and through changes in mechanoreceptor sensitivity (Mense 1993). Flor et al. (1985, 1991, 1992) proposed a “symptom-specificity” model of chronic pain based on principles of individual-response stereotypy (see Andreassi 1995). Accordingly, patients with certain kinds of musculoskeletal disorders can be distinguished from patients with other kinds of musculoskeletal disorders and from healthy controls by substantial stress-induced tension in muscles near the site of pain or injury. Chronic low back patients, for instance, can be expected to show aberrant stress-induced muscular responses specific to the disorder; that is, exaggerated contraction of the muscles in the lower back (i.e., lower paraspinals). Findings support symptom-specificity models among patients with low back pain (Arena et al. 1991; Burns et al. 1997; Flor et al. 1985, 1991, 1992, 2002; Petras and Schmidt 1991), and those with neck and shoulder pain (Lundberg et al. 1994, 1999). Indeed, Burns (2006a) reported that lower paraspinal reactivity to anger among low back patients was correlated significantly (r = .27) with everyday chronic pain severity.

With regard to anger inhibition, muscle tension reactivity and pain, we have at least two sources of evidence to link these factors. First, in two studies of low back pain patients, Burns and colleagues (Burns 1997; Burns et al. 2006) found that measures of trait anger-in and cynical hostility interacted to predict lower paraspinal reactivity during stress induction. In Burns (1997), patients with a profile of high anger-in and high hostility exhibited the greatest lower paraspinal reactivity during an anger recall interview, but not during mental arithmetic. In a replication and extension, Burns et al. (2006) found that patients characterized by this same high anger-in/high hostility profile showed the greatest lower paraspinal reactivity during an anger recall interview, but not during a sadness recall interview. Thus, effects were confined to events in which anger was aroused. Supporting a symptom-specific model, trait anger-in and hostility were not related significantly to trapezius muscle (muscles distal from the site of injury) reactivity in either study. Although trait anger-in interacted with another anger-related factor to affect lower paraspinal tension increases during anger-induction, results support the notion that at least one kind of anger inhibition is linked to muscle tension reactivity.

In sum, anger arousal among low back pain patients may lead to greater lower paraspinal tension increases than the arousal of other negative emotions, and anger-induced paraspinal reactivity may be related to everyday chronic pain severity. Patients characterized by high trait anger-in (and other factors) may be quite prone to such reactivity. Thus, not unlike the case for sympathetic nervous system reactivity, we have at this point a circumstantial case that muscle tension reactivity may mediate the link between anger inhibition and high levels of chronic pain intensity. However, we still do not know whether self-reported trait anger inhibitors, by actually regulating anger through deliberate suppression, suffer high levels of symptom-specific muscle tension reactivity, which then in turn affects subsequent chronic pain intensity.

Neural substrates underlying anger inhibition and pain

A last new direction we offer regards the growing literature examining the neural underpinnings of cognitive and emotional regulation. Using functional brain imaging techniques (i.e., fMRI), the neural basis for efforts to suppress anger and the extent to which these efforts subsequently may affect neural responses to pain can be illuminated.

Recent research involved in the cognitive regulation of emotion (e.g., reappraisal) and pain has implicated the dorsolateral (DLPFC) and ventromedial prefrontal cortices (VMPFC), as well as the anterior cingulate cortex (ACC) regions (Cacioppo et al. 2007; Ochsner and Gross 2004; Wager 2005). Evidence suggests that the ACC may subserve the attentional aspects of emotion and pain regulation, whereas it has been suggested that the DLPFC may execute regulation efforts (e.g., Fassbender et al. 2004; Ochsner 2006; Anderson et al. 2004). Indeed, there is evidence that activation in the ACC may precede activation in the DLPFC during executive tasks (Fassbender et al. 2004). These two brain regions appear to map well onto Wegner’s (1994) notions of an automatic monitoring system (i.e., ACC) that scans and attends to evidence of a mismatch between current state and intended state, and an effortful, conscious and resource-dependent operator (i.e., DLPFC) that directs thoughts to more closely match an individual’s intended state.

Indeed, Mitchell et al. (2007) found that attempts to suppress thoughts of white bears led to sustained activation in the DLPFC and transient activations in the ACC. Further, ACC activations were observed in response to occurrences of to-be-suppressed white bear thoughts. Hence, it appears that the ACC represents a conflict or error monitor. During suppression, it may seek evidence of suppression failures and so call forth action by brain regions known to subserve effective emotion regulation, such as the DLPFC. Whether these neural activation patterns hold true with respect to suppression of anger, or other emotional states, is an open question and requires further study. Nonetheless, these data provide a foundation on which to study the neural architecture of the suppression of anger-related thoughts and feelings. Specifically, during efforts to suppress anger, the ACC may act as the “monitoring process” that scans for evidence of anger (i.e., error or conflict monitoring; Fassbender et al. 2004), and subsequently signal the DLPFC as well as other emotion regulatory regions (e.g., VMPFC), to eliminate evidence of anger from conscious awareness. Interestingly, the DLPFC is a region in which semantic emotional information is stored, and so cognition pertaining to content opposite anger (e.g., positive thoughts) may be brought to bear through activation of this region in order to replace the unwanted anger-based thoughts and feelings (Barrett et al. 2007). Moreover, investigators have suggested that the DLPFC may be related to the ability to experience positive emotion (Herrington et al. 2005). Lastly, it is commonly thought that the DLPFC underlies the ability to “set-shift,” as well as other so-called executive functions likely implicated in the ability to successfully exert top-down control over emotions (Miyake et al. 2000).

To understand the neural basis of connections between anger inhibition and pain intensity, we would also be interested in neural activation patterns in response to pain-induction subsequent to active suppression. The long-term effects of emotion regulation are rarely examined in the laboratory, and investigations of links between emotion regulation at one point and responses to stress at another point are even fewer, particularly with respect to neural mechanisms. Hence, it would make sense to examine suppression versus nonsuppression differences in brain activation patterns during pain-induction in the context of our ironic process model, and the correlations between brain activation patterns during suppression and subsequent pain-induction. Candidate brain regions for investigation are the dorsal region of the ACC, which has been implicated in the affective versus sensory processing of painful stimuli, and somatosensory cortices SI and SII, as well as the pre-sensorimotor cortex (pre-SMA), which are activated by physical pain regardless of the emotional dimension (Rainville 2002). For instance, manipulations of the perceived unpleasantness of pain, keeping pain intensity constant, appear to affect the neural activation patterns of the ACC in particular, as well as subcortical and brainstem regions involved in pain processing and experience (Wager 2005; Rainville 2002; Lorenz et al. 2003), but to much lesser extent, SI, SII and pre-SMA areas. Hence, we might expect greater activation in dorsal ACC relative to SI and SII regions following efforts to suppress anger versus a control condition. We also might anticipate ACC and DLPFC activation during suppression to predict dorsal ACC activation during later pain-induction.

Although much of the foregoing is quite speculative, investigations into the neural architecture of emotion regulation and pain response are increasing in number. Tying together these levels of empirical inquiry certainly may prove useful in attempting to understand relations between state anger inhibition and pain.

Conceptualizing bases for clinical intervention

This review would be incomplete without some attention devoted to clinical implications of the relationship between anger inhibition and pain. Results strongly suggest that state anger suppression affects subsequent pain intensity, particularly the degree to which pain may be charged with negative affect. To the degree that trait anger inhibitors engage in state suppression as an anger regulation tactic, their propensity to report more chronic pain may be driven by this practice. Although evidence favors a somatic amplification explanation over an inhibition explanation for the link between trait anger-in and chronic pain, we may simply have not yet looked in all the right places. If high trait anger inhibitors do frequently engage in state anger suppression, as well as other ineffective tactics for reducing anger arousal (i.e., rumination), then clinical interventions aimed at curbing suppression or combating rumination may help alleviate chronic pain severity. Rather than simply state that trait inhibitors should disinhibit or express anger, let us set the stage for theoretical models that could guide such interventions.

Pulling together the disparate inhibition and disinhibition literatures

If state anger suppression increases subsequent pain sensitivity by making anger-related content highly accessible to consciousness (via operation of the automatic monitor), then deliberately expressing thoughts and feelings following suppression may interrupt the automatic monitor, make anger less accessible, and reduce later pain sensitivity. Evaluating such a proposition has implications for testing theory about how anger suppression affects responses to later noxious events, and how expression may prevent or even reverse these effects. If people who express following suppression not only show lower pain sensitivity compared to those who do not express, but also reveal lower anger accessibility during pain, then the notion that such ironically increased accessibility links suppressed anger to pain is greatly supported. More importantly, demonstrating that effects of state suppression can be avoided or eased through a definable expression process will shed light on therapeutic mechanisms whereby habitual anger inhibitors (or mismatched anger expressers) may benefit.

The questions become: does expression of thoughts and feelings counter the ill-effects of anger suppression on pain? and how does expression exert such effects? Evidence from the emotion regulation literature, however, does not provide clear or comprehensive answers. Results of Gross and colleagues’ indicate that expressive suppression increases SNS reactivity, and that people who habitually suppress emotion fare worse than those who typically reappraise stimuli (Gross 2002; Gross et al. 2000). The experimental paradigm used in this program, however, has not explicitly examined effects of emotion expression per se, either in comparison with suppression conditions or as a regulation tactic that may follow suppression. Thus, these findings pertain only to the “suppression-side” of regulation.

Work of Hokanson and colleagues gives some evidence that expression following anger instigation leads to swifter SNS recovery than not expressing (Hokanson and Shetler 1960). More recent work (with trait anger management style) supports the notion that expression during (Lai and Linden 1992) or following provocation does indeed facilitate SNS recovery (Engebretson et al. 1989; Faber and Burns 1996). In these studies, however, effortful state suppression (a la Wegner and Gross) by subjects during emotional arousal was not mandated. Thus, these findings pertain only to the “expression-side” of regulation.

Another literature, following Pennebaker and colleagues, purports to examine effects of expression following active suppression, and so appears to bridge the gap between suppression and expression. Here, inhibited traumatic or stressful material is expressed—verbal or written—with results suggesting that physical or psychological health benefits accrue (Boals and Klein 2005; Norman et al. 2004; Petrie et al. 1998; Smyth et al. 1999; Warner et al. 2005). This paradigm, however, differs from that of Gross, Wegner and others in that subjects in most studies have not responded to a standardized emotion stimulus, but instead have disclosed facts and feelings about their own life experiences. Thus, suppression does not occur under controlled circumstances, nor does it occur in response to a standardized stimulus. Here, only expression occurs under lab control.

On way to conceptualize the literatures on emotion regulation is that the suppression-side (Gross, Wegner, and others) has examined effects of state suppressed emotion (IV = suppress/nonsuppress) on dependent variables, and the expression-side (Hokanson, Pennebaker and others) have examined effects of emotion expression (IV = express/nonexpress) on dependent variables. We submit that these literatures are studying both sides of the “regulation coin”, but independently. Understanding of how suppression hurts and how expression prevents or undoes these deleterious effects can be improved by first studying both processes as they unfold in laboratory-based procedures, as the promising results of Liberman and Forster (2000) attest.

Moreover, examining both suppression and expression under controlled laboratory conditions can permit testing mechanisms. Catharsis, exposure, and “meaning-making” have been proposed as potential processes by which expression undoes ill-effects of suppression. A laboratory setting with control over both suppression and expression would allow determination of whether expressing negative emotions and/or attempts to make sense of an episode avoids or reduces effects of suppression. That is, only by combining the suppression-side with the expression-side of a regulatory process can we properly connect phenomena manifested during expression following state suppression (i.e., anger and/or causal words) with phenomena manifested during later pain (i.e., reduced anger accessibility; reduced pain).

Put simply, if accessibility to consciousness of anger-related content is heightened and maintained by an automatic monitoring system searching for failures to suppress the to-be-suppressed (anger) thoughts, then short-circuiting this process by encouraging these intrusions to be thought and expressed may eliminate their accessibility and influence on subsequent events. But we do not know this yet. At present, there are promising results suggesting that suppression has ill-effects on physical and psychological health, and there are some indications that expression (disclosure) following inhibition may work to reverse these effects. But the basic mechanisms—those on which we would build and refine therapeutic techniques—are far from definitive. Well-controlled studies examining whether ironic effects of anger suppression can be avoided by subsequently expressing to-be-suppressed thoughts and feelings, and whether this process of undoing ironic effects is reflected in pain variables and reduced cognitive accessibility of the to-be-suppressed content will shed light on specific causal pathways that underlie effects of suppressed anger on pain. Results of these studies may then be used as a basis to design therapeutic interventions that accurately target these empirically-supported mechanisms not only among chronic pain patients but among patients suffering from a variety of painful medical conditions.

Considering anger rumination and post-event processing

As suggested by Sukhodolsky et al. (2001), anger inhibition may be a two-stage process in which inhibition of angry behavior and speech occurs, most likely, during a provocative event, followed by some kind of rumination about the event. These two phenomena are implied by the various items of the anger-in subscale of the AEI. The items, “I keep things in,” and “I boil inside but don’t show it” appear to tap expressive suppression or behavioral inhibition, whereas the items, “I tend to harbor grudges that I don’t tell anyone about,” and “I am secretly quite critical of others” appear to tap a process akin to cognitive rumination. Thus, it appears that implicit within the most popular anger-in measure is this two-stage process of behavioral inhibition followed by cognitive rumination. There are at least two consequences of allowing trait anger-in to be defined seemly covertly in terms of these two distinct processes. The first outcome is to obscure the potentially important distinctions between behavioral anger inhibition and cognitive anger rumination. These processes may be quite separable in that, as described by Weber (2004), it is plausible that behavioral inhibition may be followed under some circumstances by anger-reducing cognitive restructuring rather than anger-amplifying rumination. Thus, some anger inhibition may not be all bad. Second, given reports of the moderate to large association between rumination and depressed mood (e.g., Nolen-Hoeksema 2000), the items of the anger-in subscale reflecting ruminative processes may account for the moderate overlap between anger-in scores and NA.

Work on rumination in depression and social anxiety disorders points to the negative effects of so-called “post-event processing” during which a certain event is rehashed with negative interpretations of self and other brought to the fore and sustained. Other work indicates that rumination involves the inability to inhibit rumination and/or to switch attention to other cognitive sets (Whitmer and Banich 2007). We propose that maladaptive anger inhibition may involve suppression of thoughts, feelings and behaviors during an anger-provoking event, followed by post-event processing and the inability to inhibit rumination. If this conceptualization is valid, then maladaptive anger inhibition may provide a harmful double dose in which suppression during provocation ironically makes anger content hyperaccessible to later, uninhibited rumination about the event. Painful stimulation experienced at some point during post-event rumination would probably be perceived in the worst light.

Although this two-stage model of anger inhibition remains speculative at present, it could help open the door to a level of clinical intervention that would complement the “anti-suppression expression” plan described above. Directly addressing post-event processing and rumination by teaching patients how to deliberately inhibit recurrent negative thoughts and/or to switch attention elsewhere, perhaps through distraction, may help interrupt the anger-inhibition—pain intensity connection.


Despite widespread acceptance of the notion, the self-reported tendency to inhibit anger expression, as measured by the very popular anger-in subscale of the AEI, may not be related reliably to acute and chronic pain intensity, at least not for the reasons we believe it to be. Our best evidence suggests that the anger-in subscale is a better measure of general NA than it is of some process of actual inhibition of anger expression that leads to increased pain sensitivity. This is not to say that the anger-in subscale does not tap the tendency to inhibit emotion expression; based on correlations with other self-report measures, it does appear to do so. With regard to the physical symptom of pain, however, the anger-in subscale seems to be related to pain mostly in common with NA. These findings could be taken to indicate that inhibited anger itself may not actually affect pain sensitivity; that nothing about the actual process of inhibiting anger expression causes pain or any other physical disorder. It may be that the link between trait anger-in and pain is due almost entirely to the hypervigilance to any sign of distress—physical or emotional—that plagues people high on NA, and thus anger-in is just a proxy for the sundry disruptions—cognitive, behavioral and physiological—that are part and parcel of chronic depression and anxiety.

We argue, however, that the conclusion that anger-in is related to pain only to the degree that is a component of NA is an artifact of the rather limited way we have studied the connection. Deriving conclusions about anger-in and pain by relying on the self-reported anger-in subscale of the AEI may not allow a thorough examination of the true effects anger inhibition may have on pain. On one level, the anger-in subscale may not sufficiently tap the full range of inhibitory processes such as thought and experiential suppression, rumination, and less-than-conscious repression, and may focus too closely on simple behavioral inhibition. Considering these additional processes may provide more comprehensive profiles of anger inhibitors, or alternatively reveal the most toxic component(s) in a multidimensional phenomenon. Regardless of the how comprehensive the profiling may be, one limitation of exclusive reliance on self-report is the method variance such assessment shares with typical measures of NA.

Thus, on another level, we believe that the manipulation of actual inhibitory phenomena in controlled laboratory settings in order to provide state assessments of effects must augment use of self-report (trait) measures. In this regard, we have found support for the notion that anger suppression—both experiential and expressive—may actually cause subsequent increased pain severity. Preliminary evidence suggests as well that the link between anger suppression and pain is specific to the process of suppression and is not immediately reducible to effects of a third variable like NA. According to an ironic process conceptualization, attempts to suppress may make anger more accessible to consciousness, thereby contributing to anger content lingering long enough to affect perceptions of later noxious events. Indeed, rebounding anger may share a close affinity to pain in particular because of subcortical ties between it and aggression.

Our recent findings for state anger suppression effects on pain cannot support definitive cause and effect inferences (yet), but results underscore two points. Despite uninspiring conclusions regarding the link between anger-in and pain based on a single trait measure, anger inhibition may indeed have unique effects on pain perception due to mechanisms inherent in attempts to inhibit thoughts, feelings and behavior, but we need to rely on well-developed theories and new research paradigms that move well-beyond self-reports of trait-like attributes. Only in this way will our understanding of the connections between inhibition and physical disorder be furthered.


Much work has been devoted to describing the construct of anger, and to attempting to differentiate anger, hostility and aggression. A thorough treatment of these efforts is beyond the scope of this review. For our purposes, we employ Spielberger et al.’s (1985) notion that anger is an emotional state that varies in intensity from mild irritation to full-blown fury. Work has also been directed toward distinguishing anger from other broad negative affects. Again, a thorough account of this research is beyond the scope of this review, and we point readers to some recent attempts to come to grips with what factors make anger a unique emotion (e.g., Harmon-Jones 2007; Harmon-Jones and Harmon-Jones 2007).



This research was supported in part by Grants MH071260 from the National Institute of Mental Health (John W. Burns, Ph.D. and Stephen Bruehl, Ph.D.), and NS046694 from the National Institute of Neurological Disorders and Stroke (Stephen Bruehl, Ph.D.).

Copyright information

© Springer Science+Business Media, LLC 2008

Authors and Affiliations

  • John W. Burns
    • 1
  • Phillip J. Quartana
    • 1
  • Stephen Bruehl
    • 2
  1. 1.Department of PsychologyRosalind Franklin University of Medicine and ScienceNorth ChicagoUSA
  2. 2.Vanderbilt University School of MedicineNashvilleUSA

Personalised recommendations